Mechanistic and ecological explanations in agent-based models of cognition